Reflection waveform inversion based on full-band seismic data reconstruction for salt structure inversion. Issue 1 (4th October 2019)

• Record Type: Journal Article
• Authors: Chen, Guoxin; Chen, Shengchang; Yang, Wencai
• Is Part Of: Geophysical journal international. Volume 220:Issue 1(2020)
• Page Start: 235
• Page End: 247
• Publication Date: 2019-10-04
• Subjects: Inverse theory -- Waveform inversion -- Acoustic properties -- Seismic tomography -- Wave propagation; Geophysics -- Periodicals; 550
• DOI: 10.1093/gji/ggz442
• Languages: English
• ISSNs: 0956-540X

Abstract

SUMMARY: Salt structures are high potential targets for oil and gas exploration. However, large-scale salt domes with irregular surfaces pose significant challenges for velocity model building. For full waveform inversion, in the absence of a high-fidelity initial model, the success of the inversion depends on low-frequency seismic data, which are scarce in the exploration data sets. This paper presents a new idea to solve the problem of salt structure velocity modelling. First, we propose an envelope-based full-band seismic data reconstruction algorithm. The smoothness of envelope is used to segment the events in seismic data, and the phase independence of envelope is used for the identification of the seismic event's arrival-time to obtain the apparent reflection sequences of the subsurface. Full-band seismic data are obtained by convolving the apparent reflection sequence with full-band source. Window averaging function and threshold strategy are used to ensure the accuracy of seismic event segmentation and the stability of the algorithm when dealing with noisy data. Then the multiscale reflection waveform inversion based on reconstructed data is proposed for salt structure velocity building. The numerical experiment results of the Sigbee2A model demonstrate the performance of the inversion algorithm in the case where the seismic data lack low-frequency components and contain noise. The limitations of the algorithm have also been analysed and studied.